Fixes parsing of hash tokens in the middle of a line.

To parse # correctly, we need to know whether it is the first token in a
line - we can deduct this either from the whitespace or seeing that the
token is the first in the file - we already calculate this information.
This patch moves the identification of the first token into the
getNextToken method and stores it inside the FormatToken, so the
UnwrappedLineParser can stay independent of the SourceManager.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@171640 91177308-0d34-0410-b5e6-96231b3b80d8

Author: r4nt

lib/Format/Format.cpp

@@ -483,8 +483,7 @@ class UnwrappedLineFormatter {
 
     unsigned Newlines =
         std::min(Token.NewlinesBefore, Style.MaxEmptyLinesToKeep + 1);
-    unsigned Offset = SourceMgr.getFileOffset(Token.WhiteSpaceStart);
-    if (Newlines == 0 && Offset != 0)
+    if (Newlines == 0 && !Token.IsFirst)
       Newlines = 1;
     unsigned Indent = Line.Level * 2;
     if ((Token.Tok.is(tok::kw_public) || Token.Tok.is(tok::kw_protected) ||
@@ -685,9 +684,10 @@ class TokenAnnotator {
       next();
       if (Index >= Tokens.size())
         return;
-      // It is the responsibility of the UnwrappedLineParser to make sure
-      // this sequence is not produced inside an unwrapped line.
-      assert(Tokens[Index].Tok.getIdentifierInfo() != NULL);
+      // Hashes in the middle of a line can lead to any strange token
+      // sequence.
+      if (Tokens[Index].Tok.getIdentifierInfo() == NULL)
+        return;
       switch (Tokens[Index].Tok.getIdentifierInfo()->getPPKeywordID()) {
       case tok::pp_include:
       case tok::pp_import:
@@ -1033,6 +1033,8 @@ class LexerBasedFormatTokenSource : public FormatTokenSource {
     Lex.LexFromRawLexer(FormatTok.Tok);
     StringRef Text = tokenText(FormatTok.Tok);
     FormatTok.WhiteSpaceStart = FormatTok.Tok.getLocation();
+    if (SourceMgr.getFileOffset(FormatTok.WhiteSpaceStart) == 0)
+      FormatTok.IsFirst = true;
 
     // Consume and record whitespace until we find a significant token.
     while (FormatTok.Tok.is(tok::unknown)) {

lib/Format/UnwrappedLineParser.cpp

@@ -470,7 +470,9 @@ void UnwrappedLineParser::nextToken() {
 
 void UnwrappedLineParser::readToken() {
   FormatTok = Tokens->getNextToken();
-  while (!Line.InPPDirective && FormatTok.Tok.is(tok::hash)) {
+  while (!Line.InPPDirective && FormatTok.Tok.is(tok::hash) &&
+         ((FormatTok.NewlinesBefore > 0 && FormatTok.HasUnescapedNewline) ||
+          FormatTok.IsFirst)) {
     // FIXME: This is incorrect - the correct way is to create a
     // data structure that will construct the parts around the preprocessor
     // directive as a structured \c UnwrappedLine.

lib/Format/UnwrappedLineParser.h

@@ -31,7 +31,8 @@ namespace format {
 /// whitespace characters preceeding it.
 struct FormatToken {
   FormatToken()
-      : NewlinesBefore(0), HasUnescapedNewline(false), WhiteSpaceLength(0) {
+      : NewlinesBefore(0), HasUnescapedNewline(false), WhiteSpaceLength(0),
+        IsFirst(false) {
   }
 
   /// \brief The \c Token.
@@ -56,6 +57,9 @@ struct FormatToken {
   /// \brief The length in characters of the whitespace immediately preceeding
   /// the \c Token.
   unsigned WhiteSpaceLength;
+
+  /// \brief Indicates that this is the first token.
+  bool IsFirst;
 };
 
 /// \brief An unwrapped line is a sequence of \c Token, that we would like to

unittests/Format/FormatTest.cpp

@@ -470,6 +470,10 @@ TEST_F(FormatTest, IndentPreprocessorDirectivesAtZero) {
   EXPECT_EQ("{\n  {\n#define A\n  }\n}", format("{{\n#define A\n}}"));
 }
 
+TEST_F(FormatTest, FormatHashIfNotAtStartOfLine) {
+  verifyFormat("{\n  {\n    a #c;\n  }\n}");
+}
+
 // FIXME: write test for unbalanced braces in macros...
 // FIXME: test # inside a normal statement (like {#define A b})